We are underway and I am long overdue in posting an entry here so there is much to discuss.
What have I been doing:
Planning! I have been getting very familiar with lighter2 and determining where change would be most appreciated. This has been a slow task as much of lighter2 is un-commented (or at least, the comments are not very detailed). Also, I have had to learn much about the CS app framework, the instance tracking classes used in CS and the i* classes used throughout lighter2. Conceptually, I reached a good place to actually propose some changes to Scott my mentor last week and we met to discuss just that.
What is the current status:
At present, we have identified the following concerns or features that need attention in lighter2 and would pertain to my proposal and my areas of expertise -
- lighter2 is an external dependency for any CS app. For redistribution purposes, it would be better if it was internal (either built into the library or as a plugin).
- lighter2 uses the 'direct lighting' approach for its central calculation. This works well and is efficient but is below the standard of other game engines which use Radiosity. For example, see this page under the heading 'Dynamic Lighting and Shadows' (http://developer.valvesoftware.com/wiki/Source_Engine_Features).
- While lighter2 is a single run, non-interactive component efficiency is still quite important. This is because during development of a game, static lighting needs to be recalculated every time the world is changed. This can be tedious if it takes more than a minute or two to give a result(and seconds would be even better).
Conceptually, any light-mapping application can be thought of as a bootstrap. The rendering system will use the light maps to render the world but the light-mapping application needs a rendering system (or at least part of one) to construct the light maps. Therefore, lighter2 naturally depends on some components of the CS library (mostly viewing and projection calculations and geometry loading components).
Some of the conceptual components of lighter2 (like the 'scene' and 'segment' classes) should be part of the CS library. I have not examined the library itself too deeply to see if it provides these components but Scott suggested that they are in fact copies of classes from the library. This is conceptually undesirable but there may be reasons for it. More discussion of this is in order.
- At the heart of the 'direct lighting' approach is a BRDF approximation used to calculate how much ambient/diffuse illumination each surface gives off for each light source. I have not yet identified where this calculation takes place but I'm operating on the assumption that this calculation is being done somewhere and is using a very simple BRDF model like the Blinn-Phong model.
What conclusions can be drawn:
From all of this I've identified some requirements for this project -
- New code should be easy to integrate (or already be integrated) into the CS library.
- The current 'lighter2' work flow (which will probably be more efficient than any new work flow) should be retained as a fallback for when quick calculations are important. This includes both the 'direct lighting' algorithm and the current BRDF model.
- We should consider re-writing lighter as a plugin. This is secondary to the main proposal objectives but if we are making enough changes to lighter this may happen anyways as a side-effect or as an add on at the end if there is extra time.
So what are the plans:
I'm going to start moving forward with changes now. Here's the initial proposal of work in the order it will be undertaken (this may change in a typical design-build fasion) -
- The direct lighting BRDF model will be upgraded to the Oren-Nayer model. This should be very easy and will satisfy part of the original proposal. If nothing else gets completed on this project at least this will be in place. It will also serve as a way to finish examining the details of lighter2 that will be more productive than just reading the source. Note that we may need to convert the Blinn-Phong parameters to Oren-Nayer parameters so that material properties do not need to be redesigned for every existing world to take advantage of this improved model.
- The Radiosity algorithm will be implemented as a CS plugin. I have implemented Radiosity before and believe this can be done relatively quickly given the tools that CS already has built-in. Implementing it as a plugin will make it available for use in lighter2 or even as its own internal step in the standard CS rendering pipeline. Here's a breakdown of how this will be undertaken -
- The most efficient approach to Radiosity that I know of is the hemi-cube method. This fits well in any scanline rendering system that can render-to-texture. We will start by developing a CS app that renders the scene from the point of view of the light sources to a texture map. This will then be integrated into a bigger radiosity plugin.
- In the radiosity algorithm all surface patches can be light sources. To avoid the need to supply seperate lights just for the radiosity system, existing light sources will have to be approximated with proxy geometry and special work may be required to support spot-lights and other non-area light sources. Point light source will need to be approximated with very small area light sources.
- The hemi-cube textures are used to compute form factors and fill in a large matrix that describes which surfaces can see what parts of which other surfaces. This is a straight-forward calculation.
- With the form factors calculated we need a large linear system solver that computes the equilibrium achieved by this world. This will initially be a standard Gauss-Seidel solver and can be upgraded to something fine-tuned to the radiosity problem at a later time.
- The solved system must be written back to the geometry by placing the computed colors into the mesh vertices which normally requires some type of interpolation. From here, we could further propagate this data out into textures which would then become the light maps for the scene.
At this point we will reevaluate and decide what is to be done next. Additional project tasks may include:
- Improving radiosity with things like: a better linear system solver, automatic geometry or light map subdivision at high-frequency artifacts and support for more types of light sources.
- Re-working of lighter2 as an internal dependency (see above)
- Other changes to lighter2 to bring it up to speed with the CS library and make it more maintainable into the future.
What's described here will constitute the bulk of this project and work will begin immediately. I intend to reevaluate progress as I go and learn more about CS and lighter2. All comments are welcome and encouraged as there are a plethora of assumptions underlying these ideas and any number of them could prove to be wrong. The collective knowledge of the CS community can do far better to identify these problems than I can digging through the mountains of code and documentation. My time is better served now making changes rather than fact-checking!
Thanks to all for reading this! I will post more as I go.